Travel planning is an interesting problem.
Almost always, there’s a bunch of subjective data in travel magazines, sometimes outdated, sometimes not, and noone has organized it yet. By organized, I actually mean structured. Structured it into something a machine can understand.
Add to this, It’s almost always a hassle and worry not knowing if you’re missing out on something.How do we organize this to be more manageable and create a recommendation system for people planning their trip?
When someone does structure this data properly and throws machines on top with proper recommendations, the result can be magical.
But how do we do this? Before we do that, lets put aside the “how-to” part of it.
- Specify that I liked <Snorkelling> (Event) at <Key West> (Place), <Art museums at New york>, <Poker at The Bellagio>. I would describe myself as <Adventerous>, <Relaxed>, <Party-animal> etc. I then want to be able to say “I’m going to Las Vegas, this is the first time I’m visiting, I have a budget of X, I have a time of Y, plan my trip”.
- This should result in an output of –
- <Breakfast at XYZ>, <Visit the Human specimen exhibit at Luxor>, <Pool at XYZ>, <Show of your choice>, <Night at XS Vegas>. (Examples, this is in no way the perfect las vegas trip.)
Well, I didn’t say it would be easy, did I?
Let me just throw out a few other random ideas that might be related to this:
- Event based check-ins – Self explanatory. You can check-into events instead of places. Might or might not be relevant. Also let the user rate each check-in at a later time. Add to the awesome-ness score.
- User generated events – Locals can offer to take someone around doing something interesting as a paid / free service. Think wine tasting, bike rides to interesting places, hikes etc.
- Some sort of algo that takes in various events, durations, “suitability/awesome-ness” scale and plugs them into the schedule. This is actually a core part of the product.
- An drag drop UI that lets you customize your trip (add or delete things you want to do) and then a mobile application side with alerts as you hit your time limits.
- MTurk from Amazon to generate a proper db of things to do at various places. Could possibly start out with a few wildly popular places and then expand from there. (I’m looking at Vegas, SF, LA, NY, Orlando). Actually, come to think of it, you might not even need MTurk for 5 flippin’ places.
- Think hooking up the recommendation system with Facebook interest graph somehow.
- Revenue model would be through Ads for “experiences”. I think that’d be a good word instead of events. Ctrl-H, replace “event” with “experience” in the article.